Testimony of Herman Jenich

National Committee on Vital and Health Statistics

Thursday, September 16, 1999

IPRO
Corporate Headquarters
Managed Care Department
1979 Marcus Avenue, First Floor
Lake Success, NY 11042-1002
516-326-7767 · 516-326-6177 [Fax]

Introduction

Good morning. My name is Herman Jenich, and I am the Associate Vice President for Managed Care at IPRO. IPRO serves as the Medicare Peer Review Organization for the State of New York, holds Medicaid oversight contracts in several states, and provides a variety of services to commercial insurers, unions, and managed care plans. Through its work with these clients, IPRO has gained extensive experience in evaluating the quality of health care, as well as the adequacy of health care data.

I am delighted to have been asked to discuss how some of the lessons that IPRO has learned can assist the National Committee on Vital and Health Statistics in its efforts to develop recommendations for how to best facilitate the development of electronic standards for Patient Medical Record Information (PMRI). As I am responsible for leading the design and implementation of a wide range of IPRO studies that evaluate health plan processes and outcomes, most of the information I will be presenting today is based on IPRO’s work in the managed care area. In general, IPRO’s work with health plans can be divided into three categories:

• HEDIS auditing: HEDIS is a set of performance measures developed and maintained by the National Committee for Quality Assurance (NCQA), and is used in purchaser evaluations of health care, NCQA’s health plan accreditation process, and consumer report cards. The 56 performance measures that currently comprise the HEDIS reporting set must generally undergo a rigorous audit process before being publicly reported. The audit process includes a detailed assessment of health plan information systems as well as a validation of medical record abstraction processes. IPRO has conducted or overseen several hundred HEDIS audits for Medicare, Medicaid, and commercial populations enrolled in health plans throughout the country.
• Encounter data validation: In recent years, HCFA and several states have directed health plans to submit member-level encounter data; however, there has been limited analytic utility of this information because of skepticism about its accuracy and completeness. IPRO has worked with several government clients to assess and improve the quality of encounter data submitted by health plans that serve Medicare and Medicaid populations.
• Quality improvement studies: IPRO has worked with health plans to implement a wide variety of clinical quality improvement projects. These projects are scientifically designed and conducted, with an emphasis on improving quality by examining patterns of care and patient outcomes.

The material that I am presenting today will draw upon IPRO’s experience with health plans in each of these three areas.

Overview

In my testimony, I will address three questions that currently confront the Committee as it develops its recommendations:

1. Why do we need comparable PMRI?
2. What are some of the system limitations that currently hinder health care data quality?
3. What are some issues that government and industry may want to address collaboratively in their development of PMRI standards?

I will spend the remainder of my time this morning providing what I hope is useful information to help the Committee answer each of these questions.

1. Why do we need comparable PMRI?

Comparable PMRI is fundamental to the ability of health plans, individual practitioners, and health care purchasers to ensure the value of health care delivered to patients. Without comparable PMRI, health care payers and providers cannot fully assess effectiveness, timeliness, access, and other important attributes of health services. For example, most health care organizations identify areas for improvement in relationship to particular goals, benchmarks, or trends. Without comparable PMRI among organizations and across time, efforts to improve the quality of care are hampered, as health plans and practitioners may draw misguided conclusions and fail to identify those areas most in need of improvement.

Currently, however, comparable PMRI is difficult and costly to obtain, since compiling full PMRI generally involves combining two sources of information: electronic administrative data and paper patient records. Health plans collect much of their electronic administrative data for purposes of billing, not performance measurement or quality improvement. Although administrative data is somewhat useful for process or outcome measurement, it is often inadequate to fully measure a health plan’s or an individual practitioner’s performance. Paper medical records have often been considered the “gold standard” of medical record information, but in many of the encounter data validation and quality improvement studies that we have conducted, we find that the medical record does not contain some of the information that has been documented in administrative data. In addition, paper records are very costly to retrieve and review, and intensive oversight must be used to ensure the validity of judgments made while abstracting data from paper records.

Even when health plans take great care to accurately process administrative data and consistently review medical records, the resulting performance rates may still not be accurate or consistent across health plans. Determining the maximum amount of acceptable bias depends, largely, on the needs of those using the data and the ability of validation techniques to detect differences among health care organizations. For example, NCQA, as part of its HEDIS Compliance AuditÔ process, requires that rates expressed as a percentage be biased by no more than 5 percentage points. While this level of specificity is generally acceptable to the users of HEDIS data, validating rates produced by multiple health plans with widely disparate systems and processes to this level of accuracy is challenging and often relies on auditor judgment rather than definitive data. The following section of this testimony describes some of these disparate systems and processes and their effect on health care data quality.

2. What are some of the system limitations that currently hinder health care data quality?

Through IPRO’s work in HEDISÒ auditing, encounter data validation, and quality improvement studies, we have identified those data system limitations that commonly preclude accurate measurement of health plan and individual practitioner performance. Below, we list and briefly describe each of these limitations. For ease of review, each limitation has been placed into one of five categories: clinical data, membership data, provider data, vendor data, and data integration.

Clinical Data (i.e., Claims/Encounter Data)

• Use of proprietary codes: Some health plans require that their providers use proprietary (i.e., home-grown) diagnosis or procedure codes in place of standard codes (e.g., CPT-4, ICD-9-CM), which hampers comparable performance measurement across health plans. In addition, individual providers working for multiple health plans may be faced with the difficulty of keeping track of multiple coding schemes.
• Modification of standard codes: Some health plans that use only standard diagnosis and procedure codes sometimes modify code definitions to accommodate billing and payment needs, which again impedes the ability to compare performance of health plans.
• Limitations of current coding systems: Current proprietary and standard coding systems, particularly those designed specifically for billing purposes, do not always capture health care data as needed for performance measurement or quality improvement purposes. Health plans sometimes need to use complex combinations of existing codes or must abstract data from paper medical records to measure their processes or outcomes.
• Use of proprietary forms: Health plans may use proprietary forms in place of standard forms (e.g., HCFA 1500, UB-92) in an effort to facilitate provider reporting of encounter data. These proprietary forms, however, often do not allow for the capture of all relevant clinical data or may not distinguish principal from secondary diagnoses and procedures.
• Variation in rigor of data editing: Health plans differ in the level of sophistication and rigor of their processes to edit and audit their claims and encounter data, which results in variation in data accuracy among health plans.
• Data in multiple, hard-to-integrate systems: Some health plans—particularly those that have recently undergone a merger, acquisition, or significant system upgrade—must merge data from multiple claims and encounter systems to accurately report performance measures.
• Data completeness problems: Many health plans—particularly those paying providers on a capitated basis—do not receive claims or encounters for all services rendered by their providers, which results in either under-reporting of true performance or the need to supplement administrative data with medical record abstraction.

Membership Data

• Lack of consistent member identifiers: Patient identifiers often vary across health plan data systems, making it difficult to consolidate all clinical, membership, and provider data for a particular measure.
• Data maintained in multiple systems: Many health plans capture enrollment data on multiple systems, particularly if they serve more than one population (Medicare, Medicaid, or commercial). Performance measurement efforts are sometimes hampered by inability to link members who transition from one population to another. Also, many health plans have difficulty tracking dependents who change their relationship to the primary subscriber.

Provider Data

• Lack of consistent provider identifiers: Provider identifiers often vary across health plan systems, making it difficult to consolidate all data required for provider profiling and quality improvement efforts. Also, even within a single system, providers may have different identification numbers for each location where they practice or for each provider group to which they belong.
• Data maintained in multiple systems: Many health plans use more than one database to house provider information. For example, provider credentialing data is often maintained separately from provider contracting and billing information. Performance measurement efforts are sometime hampered by difficulties in linking provider information that resides in multiple and sometimes incompatible databases.

Vendor Data

Health plans often rely on vendors for a wide variety of ancillary services (e.g., pharmacy, laboratory, behavioral health, home health, eye care). Vendors are sometimes also used for basic health plan functions, such as processing claims and encounters or maintaining membership or provider data. Often, however, health plans are not able to use vendor data for performance measurement purposes because:

• The vendor does not collect the required data.
• The health plan does not collect the data from vendor.
• The health plan collects vendor data but is unable to integrate the data into its own information systems.

Data Integration

For a variety of reasons mentioned above, health plans often have difficulty consolidating the claims, encounter, membership, provider, and vendor data required for performance measurement. Health plans that have multiple systems for each of these data types—as well as those that have recently undergone a merger, acquisition, or significant system upgrade—have the greatest difficulties. Each additional system from which data must be acquired increases the likelihood that the health plan will encounter incompatible coding schemes, different member and provider identifiers, and other data consolidation difficulties.

Almost all health plans are continually working to improve their ability to efficiently capture and integrate health care data that can be used for performance measurement and quality improvement activities. At this time, however, many health plans can only conduct relatively simple assessments of the effectiveness of health care if they rely solely on data that is available electronically. For comprehensive analyses of the care provided for diabetes, asthma, and other important diseases, health plans generally rely at least in part on abstraction of information from paper medical records.

3. What are some issues that government and industry may want to address collaboratively in their development of PMRI standards?

As stated several times in this testimony, health plans often cannot easily access accurate and complete PMRI. This hinders the ability of health plans and purchasers to measure outcomes, quality, and performance. The list of current system limitations is long; while some of these limitations could be addressed during the next several years, others are likely to be solved only through long-term collaboration by government agencies, employers, health plans, hospitals, individual practitioners, and quality standards organizations like NCQA and JCAHO. As part of their collaborative approach, these individuals and organizations may want to consider the following:

• The process for developing standards for electronic PMRI should be used as an opportunity to address some of the shortcomings of our current health care data infrastructure. Often, automation fails to achieve its promised potential because the new, automated systems simply codify—and sometimes even magnify—existing problems. The electronic standards should, for example, address the variations and limitations in existing coding schemes and the need for unique patient and provider identifiers for linking multiple databases.
• Any patient medical record system must consider a variety of potential users, including individual practitioners, health plans, hospitals, vendors, government and private purchasers, and even patients. A set of standards that meets the needs of only some of these constituencies is less likely to succeed than standards that address the needs of most or all.
• The success of any electronic PMRI standard relies on the commitment of those individuals who will enter data into the system. This includes not only clinicians, but also a variety of administrative staff that support clinicians in their documentation and billing efforts. Clinicians and their administrative staff are most likely to pay attention to how they enter information into a system that provides them with timely information on their performance, with comparisons to peer groups that they believe are meaningful.
• The PMRI standards development process should address how individuals will be trained to accurately code and enter medical record information. As mentioned earlier, some health plans and individual practitioners currently use non-standard coding schemes or data collection forms, and other have difficulty properly interpreting the current standard coding systems. Individual practitioners and the administrative staff that support them must fully understand PMRI coding requirements for any electronic PMRI initiative to succeed.
• The PMRI standards should include guidance on how to edit and audit electronic PMRI data. Health plans vary in the level of rigor that they use to ensure the accuracy of their administrative data. Setting minimum national standards for automated edits and system audits will help ensure the comparability of health care data across all organizations using the PMRI standards.
• The PMRI standards should be flexible to account for changes in health care delivery and information system technology. A process should be created to continuously review and improve the standards.
• Finally, but importantly, the standards for electronic PMRI should include sound requirements for data security and patient confidentiality. One unintended benefit of today’s disjointed healthcare data infrastructure is that makes it more difficult for unauthorized users of data to piece together all of the clinical information relating to a particular patient or provider. Because electronic PMRI standards will likely make it easier to link data from multiple sources, the need to protect patient and provider confidentiality will become even more important.

The development of standards for electronic PMRI will assist both health care purchasers and providers in their efforts to measure health plan performance. In addition, the PMRI standards will likely encourage the development of fully automated medical record keeping and, thereby, eliminate the need for costly abstraction from paper medical records to support performance measurement. Finally, for any electronic PMRI effort to be successful, government agencies and the private sector will both need to be active participants in the standards development and implementation process.

That concludes my comments for this morning. I look forward to responding to any questions that Committee members may have.